About me
I am a Postdoctoral Researcher at the Korteweg-de Vries Institute for Mathematics (KdVI), University of Amsterdam, working with Tim van Erven. My current research focuses on the mathematical foundations of explainable AI. My background is in sequential decision making, particularly bandit algorithms and reinforcement learning, and my broader interests include probability theory, statistical learning, convex analysis, concentration inequalities, and information theory.
I completed my PhD at the AI&ML group of Universitat Pompeu Fabra in Barcelona, under the supervision of Gergely Neu.
I have graduated from the Mathématiques, Vision, Apprentissage and the Mathématiques de l’aléatoire masters and from the maths department of École Normale Supérieure Paris.
Infos
- You can check out my work.
- You can contact me at ludovic.v.schwartz76 at gmail dot com.
News
- 02/26 : I have successfully defended my PhD at Universitat Pompeu Fabra, in front of a jury composed of Tor Lattimore, Gábor Lugosi, and Quentin Berthet.
- 12/25 : I am starting a postdoc at the Korteweg-de Vries Institute for Mathematics (KdVI), University of Amsterdam, working with Tim van Erven on the mathematical foundations of explainable AI.
- 12/25 : We are presenting two papers at NeurIPS 2025: Sparse Optimistic Information Directed Sampling and Distances for Markov chains from sample streams.
- 12/24 : We’re presenting a poster on Bisimulation Metrics are Optimal Transport Distances, and Can be Computed Efficiently at NeurIPS 2024 in Vancouver.
- 07/24 : We’re presenting a poster on Bisimulation Metrics are Optimal Transport Distances, and Can be Computed Efficiently at the Aligning Reinforcement Learning Experimentalists and Theorists workshop at ICML 2024 in Vienna.
- 07/24 : I’ve been invited to give a talk at the Foundations of RL and Control workshop at ICML 2024 in Vienna about our Optimistic Information Directed Sampling paper. Slides.
- 06/24 : I’m going to COLT 2024 to present our work on Optimistic Information Directed Sampling. Slides.
- 05/24 : I’m visiting Johannes Kirschner for 3 months at the Swiss Data Science Center. We are working on instance-dependent guarantees for regret in Sequential Decision Making for information-based algorithms.
- 07/23 : The videos of the lectures of RLSS2023 are available on YouTube.
- 06/23 : We are organizing the 2023 edition of the Reinforcement Learning Summer School in Barcelona.
- 12/22 : I’m going to NeurIPS to present our work on Lifting the Information Ratio. We extend the existing information-theoretic analysis of Thompson Sampling to the contextual bandits problem by modifying the definition of information gain and information ratio.
- 10/21 : I’m starting a PhD in RL theory with Gergely Neu.